(1) a. Es ist möglich, dass getanzt wurde
it is possible that danced was
'It is possible that there was dancing'
b. *Es ist möglich, getanzt zu werden
it is possible danced to be
Safir proposed that expletive empty categories must be governed, excluding ungoverned PRO from being expletive. A less stipulative account has recently been proposed by Tóth (2000), who invokes Speas's (1995) Principle of Economy of Projection in (2).
(2) Principle of Economy of Projection
Project XP only if XP has content, where a node X has content only if X dominates a distinct phonological matrix or a distinct semantic matrix
Tóth argues, since the head I of an infinitival clause has no independent tense or agreement content, its specifier must have independent content; a nonexpletive PRO does, but an expletive PRO does not have relevant content. A problem with this proposal is that it defeats the original object for which Speas proposed (2), namely to account for the absence of pro drop in languages with weak Agr. If an empty nonexpletive PRO can license a projection in accordance with (2), then so can an empty nonexpletive pro, regardless of the status of Agr in the language. In this talk I will argue that this problem can be avoided, while keeping the spirit of Tóth's proposal intact, if it is assumed (i) the verb's subject theta-role is always assigned to the inflectional affix on the verb, if it has one (proposed by Jelinek 1984 for polysynthetic languages, now extended to languages with poorer morphology) (ii) inflectional affixes are specified as to which person and number features they express (standard, but now also applied to infinitival affixes) (iii) in case morphology does not unambiguously indicate which suffix we are dealing with an overt subject is necessary to disambiguate; it is not so much rich agreement that licenses an empty subject, but an overt subject which licenses poor agreement, as proposed by Davis 2000 (this distinguishes pro drop from non pro drop). For (non-agreeing) infinitives only one inflectional ending is available. This has two consequences: (i) the infinitival suffix is always uniquely identified by morphology, since there are no other alternative endings; hence no overt specifier (subject) is necessary (ii) on the other hand, the lack of alternation with other affixes means that the infinitival affix's specification of the person and number features of the subject theta role it receives must be maximally underspecified: the infinitival suffix is [alpha number, beta person]. (In traditional terms: PRO is compatible with any person and number specification.) The absence of impersonal readings in infinitivals can now be explained with the help of another more or less standard assumption: affixes may not be specified disjunctively (Pinker 1984, Blevins 1995). Since the only thing [alpha number, beta person] is not compatible with is absence of such phi-features altogether, and since the infinitival affix may not be disjunctively specified as [alpha number, beta person] *or* [no phi-features], a reading in which the verb does not assign a subject theta-role to this affix (i.e. an impersonal reading) is impossible. The question then is why there can be such a thing as default agreement in finite impersonal constructions, as this would seem to require precisely this kind of disjunctive specification (the affix is either [3rd person] or [no phi-features] ('default')). However, this can be solved by regarding the relevant affix as an elsewhere case which is compatible with both readings. Suppose that features should have binary values, so that there cannot be a three-valued person feature. Instead, there are two features [speaker] and [addressee] (cf. Vainikka & Levy 1999). A finite verb's endings then can be specified as follows (ignoring the plural ones, for which an extra feature [+plur] is necessary): affix x - [+speaker]; affix y - [+addressee]; affix z - elsewhere. The 'elsewhere' statement of affix z makes it compatible with all readings in which it does not express a [+speaker] or [+addressee] role, which includes both 3rd person and impersonal readings, without having to give it a disjunctive specification (see Blevins 1995 on the fundamental distinction between disjunctive specification and elsewhere specification). (This approach hence may also explain why it seems to be 3rd person agreement that acts as default agreement in language after language). This elsewhere strategy is impossible with infinitives because there is only one affix available there and of course a form cannot be its own elsewhere case ('elsewhere' implies an opposition of at least two affixes). Hence, the infinitival affix cannot act as a default ending that does not express person/number features.
References
Blevins, James (1995). Syncretism and Paradigmatic Opposition. Linguistics and Philosophy 18, 113-152.
Davis, Henry (2000). Identifying Agreement. Glow Newsletter 44, 68-69.
Jelinek, Eloise (1984). Empty Categories, Case, and Configurationality.
Natural Language and Linguistic Theory 2, 39-76.
Pinker, Stephen (1984). Language Learnability and Language Development.
Cambridge, Mass.: Harvard University Press.
Safir, Kenneth (1985). Syntactic Chains. Cambridge: Cambridge University
Press.
Speas, Margaret (1995). Economy, Agreement and the Representation of Null
Arguments. Ms. University of Massachusetts, Amherst.
Tóth, Ildikó (2000). Inflected Infinitives in Hungarian. Ph.D.
dissertation, University of Tilburg.
Vainikka, Anne & Yonata Levy (1999). Empty Subjects in Finnish and Hebrew. Natural Language and Linguistic Theory 17, 613-671.